Section: New Results

Clustering of elementary modes and metabolic modules identification

While it is commonly admitted that metabolism is modular, the identification of metabolic modules remains an open topic. In fact, what remains open comes even upstream of any identification problem, and refers instead to the question of defining a good model for modules in metabolic networks. One would hope that such a model might enable, for instance, to automatically derive the metabolic pathways that have been discovered and painstakingly established by biochemists over the years. Elementary modes, that are informally defined as metabolic subnetworks that can function at steady state, meaning that all internal metabolites are produced and consumed in equal rates (that is, nothing accumulates internally), represent one starting point for a definition of modules that has been considered. There are two difficulties related to this however. One is that enumerating elementary modes has itself been proven (by members of the EPI) to be a hard problem, while the second is that even small networks (around 100 reaction nodes) can have millions of elementary modes. Clustering them based on, for instance, the amount of overlap, that is of shared reactions, is one idea that has been used. We attempted another definition of modules using elementary modes that is related to a form of node-covering. This has been submitted to publication. The corresponding software is available on request. This work was done in collaboration with C. E. Ferreira, E. Moreno, and P. Crescenzi.